Maneuvering target tracking with colored noise

Abstract

It is known that colored noise may degrade the performance of a tracking algorithm. A common remedy is to model colored noise as an autoregressive (AR) process and apply the measurement difference method One problem with the approach is that the AR parameters are usually unknown. In this work, we propose a new method to adaptively estimate the AR parameters. It is shown that this method is simple and practically feasible. We incorporate our method into the interacting multiple model (IMM) tracking algorithm and show that the performance is almost as good as that in the known parameters case.